Pollen transfer and patterns of reproductive success in pure and mixed populations of nectariferous Platanthera bifolia and P. chlorantha (Orchidaceae)

Study species

P. bifolia (L.) L. C. Rich. and P. chlorantha (Cust.) Rchb. are closely related and widely distributed species, which largely occur in sympatry (Hultén & Fries, 1986). In Poland, P. bifolia has a wider distribution than P. chlorantha. The studied species inhabit a wide range of habitats (Claessens & Kleynen, 2011). They overlap in flowering. The inflorescences produce 10–20 white flowers with spurs, in which nectar is secreted. P. bifolia has a narrow column, and pollinaria are parallel and close to each other, while the P. chlorantha column is wider, and the pollinaria are further apart and positioned at an angle (Nilsson, 1983, 1985; Claessens & Kleynen, 2011; Boberg et al., 2014). The caudicle in P. bifolia is much shorter than in P. chlorantha and sufficient to avoid self-pollination (Nilsson, 1983). The species also differ in terms of floral scent (Nilsson, 1983; Tollsten & Bergström, 1993; Esposito et al., 2018) and nectar chemistry (Brzosko & Bajguz, 2019). They are adapted to pollination through Lepidoptera, predominantly by sphingid and noctuid moths (Nilsson, 1983; Claessens & Kleynen, 2011). Platantherans can hybridise, and pollen transfer always takes place from a flower with a shorter spur (P. bifolia) to a flower with a longer spur (P. chlorantha) (Nilsson, 1985). Both species are predominantly allogamous (Nilsson, 1978, 1983; Tałałaj et al., 2017), but spontaneous autogamy was also noted (Brzosko, 2003).

Study area and populations

Our studies were performed in three regions of NE Poland and included populations from Białowieża National Park (BNP; two populations: BC and BF), Biebrza National Park (BbNP; one population: POG), and Suwałki Landscape Park (SLP; four populations: SMOL, LIN, BON, and POB) (all field research and laboratory experiments were approved by appropriate institution: The Environmental Ministry of Poland (DOP-WPN.286.53.2017MŚ), The Regional Director of Environmental Protection (WPN.6400.40.2016.MW, WPN.6400.15.2017.MR) and Biebrza National Park (e.g. 59a/2017)). These three areas are at a distance of 150–200 km, and have different characteristics (e.g. in climatic conditions and habitats). SLP, the most northern region, is characterised by the most severe weather conditions (lowest temperatures and shortest vegetation season). Populations existed in different habitats—from open (e.g. two meadow populations, LIN and BC, exposed more than the others to the wind and sun) to forests (Table 1). Overall, three pure P. chlorantha, two pure P. bifolia, and two mixed populations were investigated. In NE Poland, P. bifolia is more frequent, but in BNP, P. chlorantha dominates. In BbNP, only P. bifolia is present. In SLP, both species exist, and mixed populations were observed (POB1+POB2 and SMOLF). In the mixed populations POB1+POB2, P. chlorantha individuals (POB2) were less frequent and located at the border of the POB1 P. bifolia population. In the SMOLF P. bifolia population, one P. chlorantha individual was found.

All populations from the SLP were studied over 3 years (2015–2017), and in 2016, one P. chlorantha population from the BNP and one P. bifolia population from the BbNP were included to analyse flower traits. In 2017, seven populations of both species from three regions were fully investigated. Each year, all individuals from the populations were included, except for the largest SMOL P. bifolia population, which was located at the edge of the meadow and forest, partially under the tree canopy. Over 3 years, we investigated populations in forested areas, and in 2017, only 30 individuals were included from the meadow to check whether the plants’ RS depends on the habitat type (open SMOLM vs. forest SMOLF).

Levels and determinants of reproductive success

Potential reproductive barriers of two sympatric Platanthera species

Compatibility experiment, seed viability, and germination

An interspecies hand pollination experiment was conducted each year to test the compatibility between two Platanthera species. In 2015 and 2016, six P. bifolia inflorescences were experimentally pollinated using P. chlorantha pollen and nine P. chlorantha inflorescences were pollinated using P. bifolia pollen. To test the viability of seeds produced in different populations and to compare viability of seeds between the natural fruit set and fruit set from interspecies hand pollination groups, we collected the ripe fruits. We collected all fruits from hand-pollination experiments, and for natural pollination one or two capsules per shoot that were taken that were located at the lowest position on the inflorescence. Over 2 years we collected 85 fruits from P. bifolia natural pollination and 134 fruits from P. chlorantha natural pollination. From interspecific experiments, 44 fruits were collected and underwent the in vitro germination procedure. Capsules were stored at room temperature for 12 weeks and then stratified at 4 °C for 16 weeks. Before sowing, seeds were sterilised for 20 min in 5% Ca(OCl)₂ with a drop of Tween 80 (detergent) and rinsed six times in sterile distilled water. We then used a modified Malmgren’s medium for terrestrial orchids (Malmgren, 1996; Pierce et al., 2010) that was prepared according to the following protocol: agar 6 g·L⁻¹ (Biocorp, France), sucrose 10 g·L⁻¹ (Poch, Flint, MI, USA), activated charcoal 0.5 g·L⁻¹ (Sigma, St. Louis, MO, USA), (Ca)₃PO₄ 75 mg·L⁻¹, KH₂PO₄ 75 mg·L⁻¹, MgSO₄ · 7 H₂O 75 mg·L⁻¹, NH₄NO₃ 100 mg L⁻¹, NH₄H₂PO₄ 150 mg L⁻¹ (Chempur, Karlsruhe, Germany), coconut milk 50 ml·L⁻¹ (Tao Tao), and pH 5.8. Sterile media were dispensed into 50-mm diameter Petri dishes. Seeds from each capsule were inoculated onto three Petri dishes (subsamples), with approximately 300 seeds per subsample. After the Petri dishes were sealed using a double layer of Nescofilum (Bemis, USA), they were cultivated at 21 ± 2 °C in the dark for 40 days. The growth dishes were tested after incubation for progeny quality as follows: (1) seed viability including (a) the number of aborted seeds with abnormally shaped embryos or differing in colour (Goodwillie & Knight, 2006); and (b) the number of well-developed seeds with visible and swollen embryos or germinated seeds. The rate of seed germination was calculated as the proportion of germinated seeds 40 days after sowing (stage of protocorm).

Linear mixed effects models were used to test whether seed viability and germination rate differed between the species and populations using the R 3.2.4 software and lme4 package (Bates et al., 2015). Because there were usually three sets of seed sown from one fruit, an individual identification number was set as a random effect. Before analysis, the log of the germination rate was taken and seed viability was transformed using Box Cox transformation to fit the assumptions of a normal distribution. Results were reported using the sjPlot package (Lüdecke, 2018). Seed germination ratios from experimental crossings were compared using the Kruskal–Wallis test and Dunn’s test of multiple comparisons.

Levels and the determinants of reproductive success

Populations of both species differed in FRS between years (Table 2). Differences between populations of one species were sometimes larger than among distinct species. The highest fruiting among populations from the SPK over 3 years was noted in the SMOLF P. bifolia population (always almost 80% or more). Low fruiting in this region was observed in the POB1 P. bifolia and POB2 P. chlorantha populations in 2015 and 2017. In 2017, when seven populations from three regions were analysed, the highest FRS was noted in the SMOLF and SMOLM P. bifolia populations (without differences between meadow vs. forest) and in the forest P. chlorantha population (BF) from BNP, and the lowest FRS was in the P. bifolia populations (BC and POB1). The level of fruiting was not dependent on inflorescence flower position (U = 1.88, p = 0.06). Additive models of MRS and FRS outperformed models with species that interacted with variables (ΔBIC = 12.04 in FRS and ΔBIC = 11.29 in MRS). In particular for female flowers, models of RS explained the limited share of these trait variances for floral display and population density traits (Table 3). The overall FRS was not affected by the spatial structure, but it was mildly increased by floral display traits (Table 3).

MRS was significantly higher than FRS (Table 2). Among populations studied for 3 years, the highest MRS in each year was in the meadow LIN P. chlorantha population. In 2017, an extremely high pollen export was observed in two meadow populations—the LIN P. chlorantha (95.2%) and SMOLM P. bifolia (98.9%). Generally, in P. bifolia, more flowers had two pollinaria removed than in P. chlorantha (excluding the single population of each species). Temporary changes in MRS were observed in the BON and LIN P. chlorantha populations (Table 2). Pollen removal was greater in P. chlorantha and was affected both by the spatial population structure and floral display (Table 3). A denser population and higher shoots positively affected MRS, but the number of flowers had a negative effect on MRS. In each year, populations of both species differed in plant height, while temporary changes were noted only in POB1 and LIN populations (Table 2). The number of flowers per inflorescence varied significantly among populations, excluding P. bifolia in 2015 and P. chlorantha in 2016. Differences in the flower number between years were noted in POB1 and POG P. bifolia populations and three P. chlorantha populations (BON, LIN, and BF) (Table 2).

Stronger importance of floral display was noted when we compared fruiting and fruitless shoots in BC and POB1 populations with the lowest FRS. In both populations, fruitless shoots produced fewer flowers (F₁ = 11.431, p = 0.001 and F₁ = 9.373, p = 0.003) and had shorter inflorescences (F₁ = 12.224, p = 0.0007 and F₁ = 3.984, p = 0.05). In BON, there were fewer flowers on fruitless shoots than on those with fruits (F₁ = 15.486, p = 0.0002).

We found a difference in MRS:FRS ratios and the highest disparity was observed in meadow populations of both species, especially in the BC P. bifolia population (MRS:FRS equalled five, Table 2). In the remaining populations, it ranged from 1 to 2.1 in different years.

Flower traits

Both species were represented by short- and long-spurred populations. Generally, the longer spurs (over 3 cm on average in each year) were observed in P. chlorantha, excluding LIN (Table S1). We also noted one long-spurred P. bifolia population (POG). The values of this trait varied substantially between populations of the two species, excluding P. bifolia in 2015 (Table S1). Temporal differences in spur length were noted only in the BF P. chlorantha population. Labellum length and width, size of spur entrance, distance between the viscidia, and pollinarium length were species-specific, and generally had larger values in P. chlorantha (Table S1). In BON and POB2 P. chlorantha populations, we found a relatively high proportion of individuals with overlapping values of some traits with those in the P. bifolia (Fig. S1). Among populations of both species, temporary variation was observed (Table S1).

The nectar column did not differ significantly between P. chlorantha populations(F₁ = 1.436, p = 0.232), nor between P. bifolia (F₁ = 0.974, p = 0.325) in both years. A temporary difference in the nectar column was found only in LIN (F₁ = 15.371, p = 0.0001). In some populations and some years (five of 12 cases), the nectar column was positively correlated with the spur length (Table S1).

PCA showed similar results for the 2 years and explained 89.39% and 88.65% of the variance along two main axes (Fig. 1). The first principal component encompassed 75.22% in 2016 and 74.52% in 2017 of the total variance. The flower traits clearly differentiated two species into separate clusters (Fig. 1). The cluster representing P. bifolia is more coherent than that of P. chlorantha. P. chlorantha populations form more discrete groups, especially the BON population, which was separated from the others (Fig. 1). The intermediate position between P. bifolia and P. chlorantha clusters occupied putative hybrids (from 1% to 7% of the populations). They exist primarily in mixed (POB1+POB2) and in BON P. chlorantha populations. In 2016, one potential hybrid in the SMOLF P. bifolia population with one P. chlorantha individual was found. PCA results suggest the presence of one intermediate in the BF and LIN in 2017. In 2016 (more shoots were observed in the BON and POB2 P. chlorantha populations), four such individuals were positioned on the PCA diagram closer to P. bifolia than to P. chlorantha individuals from the other or even the same population of this species (Fig. 1). Both individuals present in the POB2 in 2017 showed intermediate traits. Models of RS, including species only as an additive factor, outperformed those where there was interaction with the flower traits (by ΔBIC = 22.71 in males and 10.49 in females). The models explained slightly more of the MRS variance compared to the FRS variance (19.5% vs. 16%). Both MRS and FRS were affected by flower traits and were lower in P. chlorantha (Table 4). In both sexes, the RS increased with increase in the spur entrance width, but it decreased with the distance between viscidia. FRS decreased when the pollinarium length increased.

Table 4:

Results of generalized linear models with beta distribution explaining male and female reproductive success in 2016–2017 in Platantherans in North-Eastern Poland with flower morphology traits.

Significant differences are given in bold.

	Male reproductive success			Female reproductive success
Predictors	Estimates	CI	p	Estimates	CI	p
(Intercept, incl. bifolia)	1.16	[0.80–2.80]	<0.001	1.16	[−0.02 to 2.35]	0.054
Species [chlorantha]	1.51	[1.21–1.81]	<0.001	0.96	[0.62 to −1.30]	<0.001
Spur length	−0.97	[−1.32 to −0.62]	<0.001	−1.23	[−1.65 to 0.82]	<0.001
Spur entrance	−0.16	[−0.34 to 0.01]	0.072	0.11	[−0.10 to 0.32]	0.301
Distance between viscidia	0.32	[0.07–0.57]	0.014	−0.43	[0.13–0.73]	0.005
Pollinarium length	−0.41	[−0.58 to −0.23]	<0.001	−0.40	[−0.59 to −0.20]	<0.001
Observations	237			247
R2	0.195			0.164

DOI: 10.7717/peerj.13362/table-4

Pollen flow

Pollen with pigments was transferred up to 25 m (average 8.2 ± 4.83 m) from the donor plants (Fig. 2, Fig. S1) and was found on 102, 43, 14, and six stigmas in the LIN, SMOLF, BON, and POB1 populations, respectively. During one month visit, up to nine flowers on the inflorescence may be pollinated, but most frequently, one to three flowers were visited. Pollinator-mediated autogamy or geitonogamy were observed in 13.7%, 67.4%, 78.5%, and 100% of flowers with marked pollen deposited on the stigma in the LIN, SMOLF, BON, and POB1 populations, respectively. Pollen deposited on some stigmas was derived from different sources—in the LIN population we found two colours on six plants, while in the SMOLF population we observed two colours on one shoot (Fig. S1). In the first population, stigmas on one plant possessed pollen with three different pigments. In the SMOLF population, stigmas in seven P. chlorantha flowers (one individual among P. bifolia) received red coloured pollen from P. bifolia. All seven flowers developed into fruits.

Compatibility experiments and seed viability and germination

Seed viability differed between species and their crossings (Kruskal–Wallis, χ² = 18.486, p = 0.00035), with P. chlorantha showing higher seed viability. It also differed between some populations (Fig. 3, Table S2). The seed germination rate was generally low, but it was clearly higher (p = 0.001) in both P. bifolia populations compared to any of the P. chlorantha populations (Fig. 4, Table S3).

In the crossing experiments, the germination rate differed between combinations of species and their crossings, incl. crossing direction (Kruskal–Wallis, χ² = 20.184, p = 0.00015). A significantly higher germination rate was attributed to P. bifolia than to P. chlorantha seeds, which developed as a result of interspecies crosses (Fig. 4, Table 5). We did not find differences in the rate of seed germination between P. chlorantha seeds from natural pollination and seeds from interspecies crosses when the receiver was P. chlorantha and the donor was P. bifolia. A similar pattern of seed germination was observed in both years.

Interspecies cross-pollination showed that bifolia-M/chlorantha-F crossings had high seed viability and were comparable with P. chlorantha, while bifolia-M/chloranthaF crossings and P. bifolia showed the lowest seed variability (Fig. 3, Table 5). The germination rate, however, was the highest in P. bifolia and very low in P. chlorantha and both crossings (Fig. 4, Table 5). For the crossings, the germination rate was significantly higher in bifolia-M/chlorantha-F than chlorantha-M/bifolia-F crossings.

Reproductive success and its determinants

Our results document a high level of RS of both Platantherans, excluding two P. bifolia populations. It is consistent with the rule that nectar-rewarding orchids are more successful in setting fruits than nectarless ones (Neiland & Wilcock, 1998; Tremblay et al., 2005). The RS in orchids is greatly differentiated and depends on many factors, one of which is floral display. Plants with larger inflorescences and more flowers better attract pollinators, which visit more flowers on larger inflorescences (Hodges, 1995; Maad, 2000; Grindeland, Sletvold & Ims, 2005; Vallius, Arminen & Salonen, 2006). If larger inflorescences are subject to factors that decrease their fitness (higher geitonogamy or intense herbivory), small inflorescences are favoured (Calvo, 1990; Vallius, Arminen & Salonen, 2006; Pellegrino, Bellusci & Musacchio, 2010). In our studies, the number of flowers was more important for RS than plant height, although in the BON P. chlorantha population with the lowest shoot density, the number of fruits depended on shoot height each year. The MRS was less dependent on the floral display, which contrasts with Bateman’s principle (the male function hypothesis) that selection of the flower number should be stronger through male compared to female function because fruiting is resource-limited (Burd & Callahan, 2000). Maad & Alexandersson (2004) found sex-differentiated selection of flower number in P. bifolia, but only through male function under drought conditions. The influence of pollinators on inflorescence height may be stronger in taller than in shorter vegetation (Ågren, Fortunel & Ehrlén, 2006; Sletvold, Grindeland & Ågren, 2013; Boberg et al., 2014), but our results do not support this finding.

Phenotypic selection depends on the mutual match between pollinator and flower morphology (Moré et al., 2012). Therefore, flower traits play a crucial role in attracting pollinators. In long-spurred plants, among the most important flower traits influencing RS is spur length (Inoue, 1986; Nilsson, 1988; Maad, 2000; Alexandersson & Johnson, 2002; Little, Dieringer & Romano, 2005; Whittall & Hodges, 2007; Boberg & Ågren, 2009; Boberg et al., 2014; Trunschke, Sletvold & Ågren, 2018). In our studies, the fruit number increased with the increase in spur length in P. bifolia. This is consistent with the expectation that selection on longer spurs occurs if the spur is too short to match the primary pollinator proboscis length (Maad, 2000; Boberg & Ågren, 2009; Boberg et al., 2014). In P. chlorantha, depending on the year, the spur length increased or decreased FRS. This may be caused by a change in the main pollinator between years. Nilsson (1988) noted the effect of selection on spur length when the spur is longer than the moths’ proboscis. A negative relationship between RS parameters and the distance between viscidia and the pollinarium length suggest mismatch between flower structure and pollinators. The weak correlations between flower traits and RS could be also explained by Boberg & Ågren (2009)’s suggestion that the flower properties might be less important in plants pollinated by nocturnal insects. It can also indicate that plant-pollinator interactions are more complicated than expected, and RS is the effect of multiple factors acting together. The cause of weak or a lack of such relationships may be the small population size. A stronger influence of floral display and flower traits was noted when we pooled the data and compared fruitless shoots (which were shorter, had shorter spikes and fewer flowers) with those producing fruits.

RS in nectariferous plants is greatly affected by nectar quantity and quality (Duffy & Stout, 2008; Willmer, 2011; Gijbels, Van den Ende & Honnay, 2014; Gijbels et al., 2015; Gijbels, Van den Ende & Honnay, 2015; Brzosko & Bajguz, 2019). Higher nectar availability promotes longer visitation in single flowers and all the inflorescences (Hodges, 1995; Maad & Reinhammar, 2004), and in effect increases selfing (de Jong, Waser & Klinkhamer, 1993; Jersakova & Johnson, 2006). Similarly to Ackerman, Rodriguez-Robles & Melendez (1994) and Duffy & Stout (2008), we found a minimal effect of nectar level on RS, suggesting that the nectar quantity offered by orchids in the populations studied is sufficient to attract pollinators.

Viability of seeds from natural pollination, although varied between populations, was generally low and may be explained by the realised mating system. A pollen-tracking experiment showed that autogamy or geitonogamy are important for RS of both species. Selfing may increase fruiting, but can decrease seeds viability and cause inbreeding depression (Grindeland, Sletvold & Ims, 2005; Duffy & Stout, 2008), which is higher in rewarding plants (Sletvold et al., 2012). This effect could be elevated due to the small population sizes, which cause an increase in the selfing or mating between relatives. More frequent geitonogamy in small rather than large P. bifolia and P. leucophaea populations was observed by Maad & Reinhammar (2004) and Wallace (2002). Lower seed quality as result of selfing was found by Tałałaj et al. (2017) in P. bifolia and Metsare et al. (2015) in P. chlorantha. Generally, the populations studied occupied small areas with plants growing in close vicinity. Thus, for many generations, gene exchange took place between the same pool of individuals—pollen was transported mainly for short distances, and seeds are dispersed in close vicinity to the mother plants (Brzosko et al., 2017). All these circumstances promote selfing and crosses between relatives. Such an explanation of low seeds viability due to the above-mentioned causes confirm studies on genetic variation conducted in the same populations (Brzosko et al., 2009; Brzosko & Wróblewska, 2013).

Spatial aspects of reproductive success

Plant-pollinator interactions and reproductive processes may depend on spatial distribution of resources in the population (e.g. nectar reward). It can influence pollinator activity and consequently affect the plant mating system and RS in different ways (Internicola et al., 2006; Juillet et al., 2007; Brys, Jacquemyn & Hermy, 2008; Duffy & Stout, 2008). According to the optimal foraging theory, the visitation rate is higher in dense patches, but in isolated plants, insects visit more flowers (Pyke, 1982). Duffy & Stout (2008) observed a lower visitation rate in denser patches of Spiranthes romanzoffiana due to competition for pollinators, while Grindeland, Sletvold & Ims (2005) did not find any effects caused by plant density. We noted that MRS, but not FRS, was density dependent. Higher MRS in denser patches suggests a higher moth visitation rate in these places, but the lack of correlation between shoot density and FRS may indicate that these visitors were not always pollinators. Despite the lack of density influence on FRS, we found a differentiated spatial pattern of fruiting in populations. In the POB1 and BC P. bifolia and the BON P. chlorantha populations, we noted a high proportion of fruitless shoots and low MRS. A possible cause for the high number of unvisited plants in the BON P. chlorantha population could be the low shoot density. Single shoots usually grow over long distances, causing low floral display, and thus are less visible for pollinators. This effect may be stronger in situations with a pollinator deficiency. However, P. chlorantha is almost exclusively the one species that flowers during this time and one of the tallest on the forest floor. For this reason, its visibility should be appropriate. In the BC P. bifolia population, a high proportion of fruitless shoots could be explained through competition for pollinators and/or their deficiency, as the density of both P. bifolia and co-flowering plants was high in this location.

We found that, in many cases, the number of pollinaria removed from the Platantherans flowers was much higher than the number of pollinated flowers and consequently the fruit set. Low pollination efficiency in the populations of other nectariferous species (Listera ovata) was found by Brys, Jacquemyn & Hermy (2008). According to Tremblay et al. (2005), in nectariferous temperate orchids, about 51.9% of pollinia were removed, while 41.8% of flowers set fruits. Maad (2000) noted lower MRS than FRS in P. bifolia and P. chlorantha populations. Our results show an opposite pattern than Maad’s (2000) observations in most populations, especially in BC and POB, where MRS was few times greater than FRS. The imbalance between MRS and FRS could be explained by the low efficiency of accidental visitors, which remove pollinaria but are not able to transfer them onto the stigmas due to a mismatch in partner morphology (Bateman, James & Rudall, 2012). Another explanation for this imbalance is high pollen discounting; the highest MRS:FRS ratio was found in meadow populations more exposed to wind. We observed pollinaria on neighbouring plants similarly to Friesen & Westwood (2013). Pollen discounting is also high when pollinators longer penetrate one inflorescence. Such behaviour is known for nectariferous plants (Vallius, Arminen & Salonen, 2006), and our experiment with pollen tracking confirms this finding.

Moth assemblages

Plant species evolution is driven by many factors that have different roles in space and time. One of the most important evolutionary mechanisms is plant–pollinator interactions, which are reflected in the level of RS (Cozzolino & Widmer, 2005; Tremblay et al., 2005). Both Platantherans may share main pollinators (Deilephila species and Hyloicus pinastri; Nilsson, 1983), although more recent studies show differences in the main pollinators, which depend on the location in the geographical range and plant communities (Maad, 2000; Boberg et al., 2014; Sexton, 2014; Esposito, Merckx & Tyteca, 2017; Mõtlep et al., 2018). Noctuids pollinate mainly P. chlorantha, while sphingids P. bifolia, although the short-spurred flowers of the last species are also pollinated by noctuids (Nilsson, 1983). Common visitors of P. bifolia in Sweden were Deilephila porcellus, D. elpenor, Hyloicus pinastri, Sphinx ligustri, and Hyles gallii (Nilsson, 1983, 1988), but in central Sweden it was pollinated almost exclusively by H. pinastri (Maad, 2000). In Estonia, P. bifolia is mainly pollinated by H. pinastri and S. ligustri, while P. chlorantha by Noctuidae (Mõtlep et al., 2018). Boberg et al. (2014) found a pollinator shift in Scandinavian P. bifolia populations, and depending on flower traits, habitat and the geographical location, the main pollinators were H. pinastri, S. ligustri, D. porcellus or Entephria caesiata. We found an important part of the whole set of known pollinators of both species, but pollinaria were found on only one Cucullia umbratica in the P. chlorantha LIN population (the main pollinator in Belgium; Esposito, Merckx & Tyteca, 2017). An important pollinator, D. porcellus, was absent in our studies, and only one individual of another important pollinator, the long-tongued S. ligustri, was found in the SMOLF P. bifolia population. In other studies, despite fruiting, pollinators have never been observed (Pérez-Hérnandez et al., 2011; Fox et al., 2015) or noted sporadically with a frequency that does not reflect the level of RS (Moré et al., 2012; Steen & Mundal, 2013; Sexton, 2014; Esposito, Merckx & Tyteca, 2017; Mõtlep et al., 2018). Correlation between the MRS and the number of moth species and their individuals suggests that moths that are not known as pollinators or those that play a minor role in other regions might enhance RS. Potentially, two noctuids (Agrotis exclamationis and Diarsia mendica), observed with high numbers in studied populations could also serve as pollinators. A relatively high RS compared to a low frequency of potential pollinators could be explained through their higher abundance on other nights. The similar fruit set in the bottom and upper parts of inflorescence indicates that pollinators are active during the entire flowering time. The low RS in some populations may suggest pollinator deficiency. This is consistent with the finding that pollen limitation due to pollinator deficiency is common in orchids (Inoue, 1985; Johnson & Bond, 1997; Maad & Alexandersson, 2004; Tremblay et al., 2005; Vallius & Salonen, 2006; Boberg et al., 2014; Zhou et al., 2016). The low number of fruiting shoots in the BON P. chlorantha and the BC and POB1 P. bifolia populations in 2017 (20–44.1%), together with a lower moth diversity and numbers also suggest pollinator deficiency. Mõtlep et al. (2018) found that pollinator abundance affects the fruit set of P. bifolia, but not P. chlorantha.

Our results support the finding that pollinators change according to geographic location and habitat (Boberg et al., 2014; Esposito, Merckx & Tyteca, 2017). Most potential pollinators recorded belonged to Noctuidae and were present in all populations, while sphingids were noted in four populations. Taking into account the distances between the spur entrance and nectar levels and the proboscis length of noctuids (5–21 mm; Boberg et al., 2014; Sexton, 2014), we can recognise many of them as pollinators. Even taking into consideration that the flowers in P. bifolia POG population have the longest spurs with the highest nectar columns, moths with shorter proboscises might try to reach the drops from the spur wall and pollinate the flower. The correlation between spur length and the nectar column in BON and BF P. chlorantha and BC P. bifolia populations suggests flower adaptation to pollinators with a longer proboscis. This could explain the case of the BON population (long-spurred, but with low nectar level), without long-tongued moths, wherein despite the high number of potential pollinators, a lower fruiting was noted. Boberg et al. (2014) found that short-spurred P. bifolia in grasslands was pollinated mainly by the short-proboscis D. porcellus, while long-spurred in woodlands by the long-proboscis S. ligustri. Nilsson (1983) showed that both moths with long and short tongues may compete for nectar.

The pollen-tracking experiment documented the activity of pollinators. Pollinia were transported up to 25 m (average 8.2 m). Up to nine flowers per inflorescence were visited by one moth, which confirms that on inflorescences of nectariferous species, pollinators stay longer. On some stigmas, massulaes from different individuals were deposited, which indicates that more than one pollinator can visit a single flower. We also found that pollinator-mediated selfing in both species increased the fruiting (auto- or geitonogamy we observed in 13.7%, 67.4%, 78.5%, and 100% of flowers on inflorescences in different populations).

Are reproductive isolation mechanisms enough for maintaining species integrity?

P. bifolia and P. chlorantha can produce hybrids (Nilsson, 1983; Maad & Nilsson, 2004; Bateman, James & Rudall, 2012; Esposito et al., 2018; Mõtlep et al., 2021). Our PCA analysis demonstrated the presence of intermediate morphotypes, localised mainly in mixed POB1+POB2 and in the BON P. chlorantha populations. Although we did not conduct genetic analyses, results of Mõtlep et al. (2021), who found that morphological traits well correlated with molecular traits, could suggest a high probability that the intermediate morphotypes are hybrids. The presence of putative hybrids in BON could be explained by mating between P. chlorantha from this population and P. bifolia individuals which we observed in 2005 in a neighbouring meadow, but later disappeared. One plant with intermediate traits is visible on the PCA diagram in 2016 in the SMOLF P. bifolia population, with a single P. chlorantha. Similarly to other studies (e.g., Nilsson, 1983; Esposito et al., 2018; Mõtlep et al., 2021) we found that intermediates are rare (1–7% of a particular population). Their number is probably higher because intermediates found in different years are rarely the same individuals (due to dormancy; Brzosko, 2003). In 2015, in the BON population, we observed more plants with intermediate flowers. They had pollinaria distanced as in P. bifolia, but a wide spur entrance and large stigma as in P. chlorantha. Results of experiments with interspecies crosses suggest a high potential for hybridisation. Seeds developed from interspecies crosses were viable and able to germinate, but with different frequency, which is in contrast to Mõtlep et al. (2021), who found no differences both in viable seeds and germination rate after interspecies crosses. In our studies, when P. bifolia was a receiver of P. chlorantha pollen, seeds were less viable than those from crosses in the opposite direction. Germination of both species was lower than seed viability, but the P. bifolia seeds germinated with higher frequency, which could be explained by a distinct mechanism controlling the germination of two species. However, germination was independent of the cross direction, which may suggest that postzygotic barriers act at different developmental phases. Lower seed viability from interspecies crosses with P. chlorantha as a pollen donor compared to the germination level of seeds produced naturally in P. bifolia may also suggest the presence of post-pollination isolation barriers. Spontaneous interspecies crossing was only noted in the SMOLF population, where flowers on only one P. chlorantha were pollinated by coloured P. bifolia pollen, and all of them set fruits with viable seeds. It confirms the Nilsson (1985) and Mõtlep et al. (2021) observations that pollen transfer always takes place from P. bifolia to P. chlorantha. This also shows that the same moths can pollinate both species. Nilsson (1992) found that the intermediate column alters interactions with moths, and consequently RS decreases. We confirm this finding; in 93% of cases, FRS and MRS of intermediates was lower. Our results could also suggest that intermediates are characterised by worse survival, because in populations with the highest number of intermediates, we observed a drastic decrease in individuals from year to year.